Characterization of Porous Building Materials for Agent Transport Predictions Using Artificial Neural Networks

Agent fate in porous building materials is currently a topic of national concern, with possible applications ranging from chemical and biological attacks to more common acid rain events. Paramount to understanding such transport phenomena is knowledge of both micro and macroscopic substrate properties such as pore structure, surface and macroscopic gas permeability, hydraulic conductivity and effective porosity, among others. In order to quickly identify and asses a traumatic situation, especially when historic and significant buildings are affected, one needs to rapidly determine such material properties using only what is available; namely the surface of the material. A study evaluating transport properties of natural and man-made porous building materials is presented. The objectives of the study were to: (1) determine substrate material properties, specifically surface gas permeability, and relate that to agent penetration and (2) develop artificial neural networks (ANNs) that will (a) determine the minimum number of sampling points needed to accurately estimate the surface permeability field and (b) predict the penetration depth of chemical warfare (via the use of surrogates) within building substrates. Surface and macroscopic permeability measurements were determined on materials ranging from concretes, to sandstones, to brick, with permeability values within the range of values reported in the literature. Wicking experiments were also performed on all the building materials. These tests provided information about the height of penetration that a surrogate solution travels in a substrate as well as the total volume of the solution wicked. Due to the heterogeneities across the substrates, it was found that both the height and volume of solution wicked varied greatly among the materials tested. The ANN methods produced results that were statistically similar to traditional geostatistical methods (ordinary kriging) and that, based on known measurements, predicted the transport depth of the surrogate solutions